Surrogate Global Optimization for Identifying Cost?Effective Green Infrastructure for Urban Flood Control With a Computationally Expensive Inundation Model

Optimization algorithms and urban inundation models are powerful tools to identify cost-effective designs of green infrastructures such as low-impact developments (LIDs). Most previous LID design optimization studies based on one-dimensional (1D) models, which cannot provide spatial information flooding. The two-dimensional (2D) or coupled 1D-2D was rarely explored due the expensive computing time. This work investigates effectiveness surrogate methods for design, have not been used problems. We propose a general framework that searches optimal configurations both flood damage under series probable events life cycle costs LID. demonstrate using case study an catchment with 55 sub-catchments 103 decision variables. tested two different designed high-dimensional problems: DYnamic COordinate search Response Surface (DYCORS) Trust Region Bayesian (TuRBO), one popular non-surrogate method particle swarm (PSO). result indicates that: (1) DYCORS is promising (significantly faster than TuRBO PSO) identifying minimize cost cost; (2) Optimized could reduce by much $12.14 million after eliminating its own compared no implementation; (3) effective in reducing imperviousness lands areas.